using 
CSV,
Dates,
DataFrames,
Format,
Markdown,
Measures,
Plots,
RCall,
Statistics,
StatsPlots,
Suppressor
@suppress R"library(forecast)"
@suppress gr()
;

death_path = "COVID-19/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_deaths_global.csv"
confirmed_path = "COVID-19/csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_confirmed_global.csv"
;

death = CSV.read(death_path) |> DataFrame! |> x -> rename!(x, Dict(Symbol("Province/State") => :state, Symbol("Country/Region") => :country))
confirmed = CSV.read(confirmed_path) |> DataFrame! |> x -> rename!(x, Dict(Symbol("Province/State") => :state, Symbol("Country/Region") => :country))
;

"""
    convert an array of columns and column names into a markdown table

    parameters: 
        cols : Array[Array[Any]] | DataFrame
            an array of arrays which are the columns of the table
        names : Array[String]
            an array of column names. If inputting a df, input `string.(names(df))`
        df : Boolean
            whether or not the cols are an array of arrays or a dataframe

    returns: String
        string of markdown text
"""
function markdown_table(cols, names; df=false)
    if df
        cols = [col for col in eachcol(cols)]
    end
    output_string = ""
    cols = [repr.(col) for col in cols]
    for i in 1:length(names)
        output_string = output_string * "|" * names[i]
    end
    output_string *= "\n"
    for i in 1:length(names)
        output_string = output_string * "|" * "---"
    end
    output_string *= "\n"
    for row in 1:length(cols[1])
        for col in 1:length(cols)
            output_string = output_string * "|" * cols[col][row]
        end
        output_string *= "\n"
    end
    # displaying strings without quotes
    output_string =  replace(output_string, "\"" => "")

    return output_string
end
;

function country_data(country; state=false, start_date=Dates.Date(2020, 1, 22), end_date="max")
    aggregate(x) = [sum(col) for col in eachcol(x)]
    
    state_mask = typeof(state) == String ? death.state .== state : true
    d = death[(death.country .== country) .& state_mask, 5:end] |> aggregate
    dates = [Dates.Date(2020, 1, 22) + Dates.Day(day) for day in 1:length(d)] |> x -> reshape(x, :, 1)
    state_mask = typeof(state) == String ? confirmed.state .== state : true
    c = confirmed[(confirmed.country .== country) .& state_mask, 5:end] |> aggregate
    
    df = convert(DataFrame, dates) |> x -> rename!(x, [:date])
    country_name = typeof(state) == String ? country * "-" * state : country
    df[!, :country] .= country_name
    df[!, :confirmed] = c
    df[!, :deaths] = d
    df[!, :death_rate] = d ./ c
    df[!, :new_cases] .= 0
    df[2:end, :new_cases] = df[2:end, :confirmed] .- df[1:end-1, :confirmed]
    df[!, :new_deaths] .= 0
    df[2:end, :new_deaths] = df[2:end, :deaths] .- df[1:end-1, :deaths]
    df[!, :acceleration_cases] .= 0
    df[2:end, :acceleration_cases] = df[2:end, :new_cases] .- df[1:end-1, :new_cases]
    df[!, :acceleration_deaths] .= 0
    df[2:end, :acceleration_deaths] = df[2:end, :new_deaths] .- df[1:end-1, :new_deaths]
    df[!, :days_since_100] .= 0
    counter = 0
    for row in eachrow(df)
        if row.confirmed > 100
            counter += 1
        end
        row.days_since_100 = counter
    end
    df[!, :days_since_10] .= 0
    counter = 0
    for row in eachrow(df)
        if row.deaths > 10
            counter += 1
        end
        row.days_since_10 = counter
    end
    
    if end_date == "max"
        end_date = maximum(df.date)
    end
    return df[(df.date .>= start_date) .& (df.date .<= end_date), :]
end
;

function plot_country(country; state=false, metric=:confirmed, start_date=Dates.Date(2020, 1, 22), end_date="max", days_since_100=false)
    data = country_data(country; state=state, start_date=start_date, end_date=end_date)
    if days_since_100
        if metric == :confirmed
            plot(data[data.days_since_100 .> 0, :days_since_100], data[data.days_since_100 .> 0, metric], label=country, legend=:outertopright, size=(1000, 500))
        else
            plot(data[data.days_since_10 .> 0, :days_since_10], data[data.days_since_10 .> 0, metric], label=country, legend=:outertopright, size=(1000, 500))
        end
    else
        plot(data.date, data[!, metric], label=country, legend=:outertopright, size=(1000, 500))
    end
    
    plot!([0], linetype=:hline, color=:black, label="")
    plot!(yformatter=:plain)
end
function plot_country!(country; state=false, metric=:confirmed, start_date=Dates.Date(2020, 1, 22), end_date="max", days_since_100=false)
    data = country_data(country; state=state, start_date=start_date, end_date=end_date)
    if days_since_100
        if metric == :confirmed
            plot!(data[data.days_since_100 .> 0, :days_since_100], data[data.days_since_100 .> 0, metric], label=country, legend=:outertopright, size=(1000, 500))
        else
            plot!(data[data.days_since_10 .> 0, :days_since_10], data[data.days_since_10 .> 0, metric], label=country, legend=:outertopright, size=(1000, 500))
        end
    else
        plot!(data.date, data[!, metric], label=country, legend=:outertopright, size=(1000, 500))
    end
    plot!(yformatter=:plain)
end
bar_plot(country, metric) = bar(
    all_country_data[
        (all_country_data.country .== country) .& (all_country_data.date .>= Dates.Date(2020, 1, 1)),
        :date
    ], 
    all_country_data[
        (all_country_data.country .== country) .& (all_country_data.date .>= Dates.Date(2020, 1, 1)),
        metric
    ], 
    legend=false, 
    linecolor=1, 
    title=country,
    yformatter=:plain
)
function accel_plot(country_name="World")
    if country_name == "World"
        tmp = all_country_data
    else
        tmp = all_country_data[all_country_data.country .== country_name, :]
    end
    agg = by(tmp, [:date], :confirmed => sum, :deaths => sum)
    conf = plot(agg.date, agg.confirmed_sum, legend=false, title="$country_name Confirmed Cases", size=(1400, 700), color=:grey, yformatter=:plain)
    death_plot = plot(agg.date, agg.deaths_sum, legend=false, title="$country_name Deaths", size=(1400, 700), color=:orange, yformatter=:plain)

    # derivates
    agg[!, :new_cases] .= 0
    agg[!, :new_cases][2:end] = agg.confirmed_sum[2:end] - agg.confirmed_sum[1:end-1]
    agg[!, :new_deaths] .= 0
    agg[!, :new_deaths][2:end] = agg.deaths_sum[2:end] - agg.deaths_sum[1:end-1]
    agg[!, :conf_acceleration] .= 0
    agg[!, :conf_acceleration][2:end] = agg.new_cases[2:end] - agg.new_cases[1:end-1]
    agg[!, :death_acceleration] .= 0
    agg[!, :death_acceleration][2:end] = agg.new_deaths[2:end] - agg.new_deaths[1:end-1]

    # moving average
    moving_average(vs,n) = [sum(@view vs[i:(i+n-1)])/n for i in 1:(length(vs)-(n-1))]
    n_days = 3
    ma_date = agg.date[1 + n_days-1:end]
    conf_slope = moving_average(agg.new_cases, n_days)
    death_slope = moving_average(agg.new_deaths, n_days)
    conf_acc = moving_average(agg.conf_acceleration, n_days)
    death_acc = moving_average(agg.death_acceleration, n_days)

    p_conf_slope = plot(ma_date, conf_slope, legend=false, title="$country_name 1st Derivative (new cases)", size=(1400, 700), color=:grey)
    plot!([0], linetype=:hline, color=:black, label="", yformatter=:plain)

    p_death_slope = plot(ma_date, death_slope, legend=false, title="$country_name 1st Derivative (new deaths)", size=(1400, 700), color=:orange)
    plot!([0], linetype=:hline, color=:black, label="", yformatter=:plain)

    p_conf_acc = plot(ma_date, conf_acc, legend=false, title="$country_name 2nd Derivative (cases acceleration)", size=(1400, 700), color=:grey)
    plot!([0], linetype=:hline, color=:black, label="", yformatter=:plain)

    p_death_acc = plot(ma_date, death_acc, legend=false, title="$country_name 2nd Derivative (deaths acceleration)", size=(1400, 700), color=:orange)
    plot!([0], linetype=:hline, color=:black, label="", yformatter=:plain)

    plot(conf, death_plot, p_conf_slope, p_death_slope, p_conf_acc, p_death_acc, layout=(3,2))
end
;

function r_forecast(x, y; r_forecast_function, time_function, h)    
    fc_x = [x[end] + time_function(i) for i in 1:h]
    fc = r_forecast_function(y)
    point_forecast = [val for val in fc[2]]
    lo_80 = [val for val in fc[6][1:h]]
    hi_80 = [val for val in fc[5][1:h]]
    lo_95 = [val for val in fc[6][h+1:h*2]]
    hi_95 = [val for val in fc[5][h+1:h*2]]
    output = hcat([point_forecast, lo_80, hi_80, lo_95, hi_95]...) |> DataFrame |> x -> rename!(x, [:point_forecast, :lo_80, :hi_80, :lo_95, :hi_95])

    p = plot(x, y, legend=false, label="Observed", color=:black, formatter=:plain)
    plot!(fc_x, point_forecast, ribbon=(point_forecast.-lo_95, hi_95.-point_forecast), label="", color=:lightblue, fillalpha=0.5)
    plot!(fc_x, point_forecast, ribbon=(point_forecast.-lo_80, hi_80.-point_forecast), label="Forecast", color=:blue, fillalpha=0.3)
    
    return [fc, output, p]
end
;

function individual_country(country; comparison_country="Italy")
    # line plots
    cases_100 = plot_country(country, days_since_100=true, metric=:confirmed)
    plot_country!(comparison_country, days_since_100=true, metric=:confirmed)
    plot!(title="$country Days Since 100th Case", titlefont=font(10))
    deaths_10 = plot_country(country, days_since_100=true, metric=:deaths)
    plot_country!(comparison_country, days_since_100=true, metric=:deaths)
    plot!(title="$country Days Since 10th Death", titlefont=font(10))
    line_plots = plot(cases_100, deaths_10, layout=(2,1))
    
    #accel plots
    accel_plots = accel_plot(country)
    plot!(titlefont=font(10))
    
    # bar plots
    cases_bar = bar_plot(country, :new_cases)
    plot!(title="$country New Cases Per Day", titlefont=font(10))
    deaths_bar = bar_plot(country, :new_deaths)
    plot!(title="$country Deaths Per Day", titlefont=font(10))
    bar_plots = plot(cases_bar, deaths_bar, layout=(2,1))
    
    # forecast plots
    data = country_data(country) |> x -> x[x.confirmed .>= 100,:]
    forecast_plot_cases = r_forecast(data.date, data.confirmed, r_forecast_function = z -> R"holt"(z, h=20, damped=true), time_function=Dates.Day, h=20)[3]
    plot!(title="$country Forecast Cases", titlefont=font(10))
    forecast_plot_deaths = r_forecast(data.date, data.deaths, r_forecast_function = z -> R"holt"(z, h=20, damped=true), time_function=Dates.Day, h=20)[3]
    plot!(title="$country Forecast Deaths", titlefont=font(10))
    forecast_plots = plot(forecast_plot_cases, forecast_plot_deaths, layout=(1, 2))
    
    final_plot = plot(accel_plots, line_plots, bar_plots, forecast_plots, layout=(4, 1))
    plot!(size=(900, 2000), margin=5mm)
    return final_plot
end
;

all_countries = death.country |> unique |> sort
all_country_data = vcat([country_data(country) for country in all_countries]...)
current_state = by(all_country_data, [:country], :confirmed => maximum, :deaths => maximum)
current_state[!, :death_rate] = current_state.deaths_maximum ./ current_state.confirmed_maximum
;

Markdown.parse("""
    # Overview of COVID-19 as of $(maximum(all_country_data.date))
    This page uses data from https://github.com/CSSEGISandData/COVID-19 (Johns Hopkins CSSE is the original source) to create a report on the status of COVID-19 cases and deaths around the world. It is updated once a day around 10:00am CET.
    """)

Worldwide Current State¶

tmp = current_state
Markdown.parse("""
- Confirmed cases: **$(format(sum(tmp.confirmed_maximum), commas=true))**
- Deaths: **$(format(sum(tmp.deaths_maximum), commas=true))**
- Death Rate: **$(round(sum(tmp.deaths_maximum) / sum(tmp.confirmed_maximum) * 100, digits=2))%**
""")

accel_plot("World")
plot!(size=(900, 700))

Top 10 Worst Affected Countries¶

Confirmed¶

top_10 = sort(current_state, order(:confirmed_maximum, rev=true))[1:10, :]
top_10[!, :death_rate] = string.(round.(top_10.death_rate * 100, digits=2)) .* "%"
Markdown.parse(
    markdown_table(top_10, ["Country", "Confirmed Cases", "Deaths", "Death Rate"]; df=true)
)

Acceleration (>1000 cases now)¶

The Acceleration of Last 5 Days column is calculated by the average second derivative over the last 5 days / number of cases 5 days ago. It doesn't have much intrinsic meaning but is rather a more comparable/relative measure between countries of how fast new cases are accelerating.

countries = unique(all_country_data.country)
acceleration = []
cases_5_ago = []
cases_now = []
for country in countries
    metric = mean(all_country_data[all_country_data.country .== country, :acceleration_cases][end-4:end])
    metric /= all_country_data[all_country_data.country .== country, :confirmed][end-4]
    if isnan(metric)
        metric = 0
    end
    push!(acceleration, metric)
    push!(cases_5_ago, all_country_data[all_country_data.country .== country, :confirmed][end-4])
    push!(cases_now, all_country_data[all_country_data.country .== country, :confirmed][end])
end
last_5 = rename!(DataFrame([countries, acceleration, cases_now, cases_5_ago]), [:country, :last_5_accel, :cases_now, :cases_5_ago])
last_5[!, :perc_increase] = last_5.cases_now ./ last_5.cases_5_ago .- 1
tmp = sort(last_5[(last_5.cases_5_ago .> 20) .& (last_5.cases_now .> 1000), :], order(:last_5_accel, rev=true))
tmp[!, :perc_increase] = string.(round.(tmp.perc_increase .* 100, digits=2)) .* "%"
rename!(tmp, [:Country, Symbol("Acceleration of Last 5 Days"), Symbol("Cases Now"), Symbol("Cases 5 Days Ago"), Symbol("% Increase in 5 Days")])
tmp[!, Symbol("Acceleration of Last 5 Days")] = string.(round.(tmp[!, Symbol("Acceleration of Last 5 Days")] * 100, digits=2)) .* "%"

Markdown.parse(
    markdown_table(tmp, string.(names(tmp)), df=true)
)

Countries¶

Days since 100th case and 10th death plotted with Italy for comparison. Only countries with >= 500 confirmed cases show graphs. Derivative plots are 3 days moving average.

Forecasts¶

Forecast numbers and plots are created using Holt's linear trend method with dampening. Important to note is that this is just one of many methods that can forecast the data, and I have not spent a significant amount of time validating it. I may spend more time in the future investigating better forecasting methods.
Forecast numbers are given for 10 and 20 days out with the 80% prediction interval. Plots display the point forecast and 80% prediction interval (dark blue shading), and 95% prediction interval (light blue shading).

for coun in all_countries
    data = country_data(coun)
    display(Markdown.parse("## $coun"))
    Markdown.parse("""
- Confirmed cases: **$(format(maximum(data.confirmed), commas=true))**
- Deaths: **$(format(maximum(data.deaths), commas=true))**
- Death Rate: **$(round(maximum(data.deaths) / maximum(data.confirmed) * 100, digits=2))%**
""") |> display
    if maximum(data.confirmed) >= 500 || coun == "Hungary" 
        h = 20
        raw_fc_data = data[data.confirmed .>= 100, :]
        fc_data_cases = r_forecast(raw_fc_data.date, raw_fc_data.confirmed, r_forecast_function = z -> R"holt"(z, h=h, damped=true), time_function=Dates.Day, h=h)[2]
        fc_data_cases = round.(fc_data_cases, digits=0)
        fc_data_deaths = r_forecast(raw_fc_data.date, raw_fc_data.deaths, r_forecast_function = z -> R"holt"(z, h=h, damped=true), time_function=Dates.Day, h=h)[2]
        fc_data_deaths = round.(fc_data_deaths, digits=0)
        _10_days = maximum(data.date) + Dates.Day(10)
        _20_days = maximum(data.date) + Dates.Day(20)
        Markdown.parse("""
**$coun Forecast**
<br>
_By $(_10_days)_
- Confirmed cases: **$(format(fc_data_cases[10, :point_forecast], commas=true))** (+/- **$(format(fc_data_cases[10, :point_forecast] - fc_data_cases[10, :lo_80], commas=true))**)
- Deaths: **$(format(fc_data_deaths[10, :point_forecast], commas=true))** (+/- **$(format(fc_data_deaths[10, :point_forecast] - fc_data_deaths[10, :lo_80], commas=true))**)
_By $(_20_days)_
- Confirmed cases: **$(format(fc_data_cases[20, :point_forecast], commas=true))** (+/- **$(format(fc_data_cases[20, :point_forecast] - fc_data_cases[20, :lo_80], commas=true))**)
- Deaths: **$(format(fc_data_deaths[20, :point_forecast], commas=true))** (+/- **$(format(fc_data_deaths[20, :point_forecast] - fc_data_deaths[20, :lo_80], commas=true))**)
""") |> display
        display(individual_country(coun))
    end
end

┌ Warning: RCall.jl: Warning in ets(x, "AAN", alpha = alpha, beta = beta, phi = phi, damped = damped,  :
│   Not enough data to use damping
└ @ RCall /Users/danielhopp/.julia/packages/RCall/g7dhB/src/io.jl:113
┌ Warning: RCall.jl: Warning in ets(x, "AAN", alpha = alpha, beta = beta, phi = phi, damped = damped,  :
│   Not enough data to use damping
└ @ RCall /Users/danielhopp/.julia/packages/RCall/g7dhB/src/io.jl:113
┌ Warning: RCall.jl: Warning in ets(x, "AAN", alpha = alpha, beta = beta, phi = phi, damped = damped,  :
│   Not enough data to use damping
└ @ RCall /Users/danielhopp/.julia/packages/RCall/g7dhB/src/io.jl:113
┌ Warning: RCall.jl: Warning in ets(x, "AAN", alpha = alpha, beta = beta, phi = phi, damped = damped,  :
│   Not enough data to use damping
└ @ RCall /Users/danielhopp/.julia/packages/RCall/g7dhB/src/io.jl:113

Country	Confirmed Cases	Deaths	Death Rate
US	213372	4757	2.23%
Italy	110574	13155	11.9%
Spain	104118	9387	9.02%
China	82361	3316	4.03%
Germany	77872	920	1.18%
France	57749	4043	7.0%
Iran	47593	3036	6.38%
United Kingdom	29865	2357	7.89%
Switzerland	17768	488	2.75%
Turkey	15679	277	1.77%

Country	Acceleration of Last 5 Days	Cases Now	Cases 5 Days Ago	% Increase in 5 Days
India	8.94%	1998	987	102.43%
Peru	6.05%	1323	671	97.17%
Russia	3.86%	2777	1264	119.7%
Colombia	3.65%	1065	608	75.16%
Brazil	3.52%	6836	3904	75.1%
Ecuador	3.47%	2748	1823	50.74%
Serbia	2.64%	1060	659	60.85%
Philippines	2.44%	2311	1075	114.98%
Dominican Republic	2.28%	1284	719	78.58%
Israel	2.17%	6092	3619	68.33%
Japan	1.7%	2178	1693	28.65%
United Kingdom	1.68%	29865	17312	72.51%
Sweden	1.64%	4947	3447	43.52%
Canada	1.41%	9560	5576	71.45%
US	1.21%	213372	121478	75.65%
Saudi Arabia	1.08%	1720	1203	42.98%
Poland	0.92%	2554	1638	55.92%
Denmark	0.63%	3290	2366	39.05%
France	0.56%	57749	38105	51.55%
Greece	0.51%	1415	1061	33.36%
Thailand	0.47%	1771	1245	42.25%
Portugal	0.32%	8251	5170	59.59%
Mexico	0.31%	1215	717	69.46%
Belgium	0.31%	13964	9134	52.88%
Turkey	0.21%	15679	7402	111.82%
Pakistan	0.11%	2118	1495	41.67%
Malaysia	0.1%	2908	2320	25.34%
Spain	0.07%	104118	73235	42.17%
Switzerland	0.07%	17768	14076	26.23%
Iran	0.04%	47593	35408	34.41%
Korea, South	0.02%	9887	9478	4.32%
China	-0.01%	82361	81999	0.44%
Iceland	-0.06%	1220	963	26.69%
Indonesia	-0.07%	1677	1155	45.19%
Chile	-0.12%	3031	1909	58.77%
Luxembourg	-0.12%	2319	1831	26.65%
Australia	-0.16%	4862	3640	33.57%
Italy	-0.24%	110574	92472	19.58%
Germany	-0.3%	77872	57695	34.97%
Netherlands	-0.31%	13696	9819	39.48%
Austria	-0.52%	10711	8271	29.5%
Romania	-0.66%	2460	1452	69.42%
Ireland	-0.75%	3447	2415	42.73%
Norway	-0.82%	4863	4015	21.12%
Finland	-0.94%	1446	1167	23.91%
Czechia	-1.17%	3508	2631	33.33%
Argentina	-2.52%	1054	690	52.75%
Panama	-2.95%	1181	786	50.25%
South Africa	-3.64%	1380	1187	16.26%

Overview of COVID-19 as of 2020-04-02¶

Worldwide Current State¶

Top 10 Worst Affected Countries¶

Confirmed¶

Acceleration (>1000 cases now)¶

Countries¶

Forecasts¶

Afghanistan¶

Albania¶

Algeria¶

Andorra¶

Angola¶

Antigua and Barbuda¶

Argentina¶

Armenia¶

Australia¶

Austria¶

Azerbaijan¶

Bahamas¶

Bahrain¶

Bangladesh¶

Barbados¶

Belarus¶

Belgium¶

Belize¶

Benin¶

Bhutan¶

Bolivia¶

Bosnia and Herzegovina¶

Botswana¶

Brazil¶

Brunei¶

Bulgaria¶

Burkina Faso¶

Burma¶

Burundi¶

Cabo Verde¶

Cambodia¶

Cameroon¶

Canada¶

Central African Republic¶

Chad¶

Chile¶

China¶

Colombia¶

Congo (Brazzaville)¶

Congo (Kinshasa)¶

Costa Rica¶

Cote d'Ivoire¶

Croatia¶

Cuba¶

Cyprus¶

Czechia¶

Denmark¶

Diamond Princess¶

Djibouti¶

Dominica¶

Dominican Republic¶

Ecuador¶

Egypt¶

El Salvador¶

Equatorial Guinea¶

Eritrea¶

Estonia¶

Eswatini¶

Ethiopia¶

Fiji¶

Finland¶

France¶

Gabon¶

Gambia¶

Georgia¶

Germany¶

Ghana¶

Greece¶

Grenada¶

Guatemala¶

Guinea¶

Guinea-Bissau¶

Guyana¶